More than one audio codec on a PC mobo

That's the brute force approach where you end up having to build everything and write the drivers yourself.

Thanks! That is very interesting and there are even expansion modules if

Thanks, Matt, that is good input. Yes, it is beyond budget but it shows that it can be done. Plus we could buy a system from Motu as a testbed.

That looks like a good solution for a testbed. Matt had a link to Motu where you can get 24 channels right off the bat, plus expansion modules.

It's not a one-off but for a product. Can't talk about the application but essentially it's the processing of electrical signals that (luckily) happen to be spectrally in the audio band. Phase synchronicity of all channels to each other and dynamic range are the key parameters.

Not really because that's just outputs. At least all the ones I've found never had more than two channels input.

If it has to be sample accurate, I think you'll want to distribute the word clock on coax from the master device to the slave analog input boxes as well. There's a switch on the back of the Behringer boxes to recover the clock from the ADAT signal, instead of using the BNC word clock input, but IIRC it could be "sloppy."

With that many, the problem is getting enough TDM channels. Most DACs only do TDM-8 so that's at least four or five channels. You're also starting to talk about a significant amount of data. An FPGA might be in order. What sampling rate do you need? 48kHz? 96kHz?

Four-channel CODECs are readily available. IIRC, six channel are around, too. I find separating them to be easier, though if they're all line level, perhaps not.

a good start would be to feed them all from the same oscillator, that should take care of the hardest problem, getting the sample rate exactly the same though there might a pll for some sample rates..

-Lasse

PLLs aren't necessary. This isn't a huge problem at all. Digital systems tend to be synchronous. ;-)

There is the PCM3168 with 6-ch input, maybe that can be coaxed into a synchronous orchestra with others.

The level can be handled via amplifiers, that's the easy part.

Feding the same clock is easy but something must make sure that the samples are all kept in time-sync. Even when they dump their data in sccessive order.

Windows and linux has whatever secret sauce is needed to make those PCI bus extenders built right in. [Actually OSX as well, but I own no Apple gear so I can't say first hand that it works on OSX.] All you have to do is make sure the cage is powered before the PC. I really wish somebody would do an open source project on these PCI bus extenders since they are horribly overpriced. Worse than GPIB!

Thanks to all the ridiculous flavors of PCI these days, you don't get a lot of PCI ports, well at least of the same type. The old PCI speed is fine for a number of audio cards.

There are hacks you can do in Linux to make the cards start up in the same order. That is, it can get confusing if every time you boot, the audio inputs have reassigned themselves. I don't know about windows.

Digidesign/Avid uses the Magma PCI gear I mentioned, well except for the name on the box. They just add other software.

Of course I meant without extra cages abnd backplanes :-)

It all goes towards USB which has better standardization. USB 3.0 has plenty of bandwidth for a gazillion audio channels. But it seems the audio world is slow to embrace that. The RF world isn't:

Den torsdag den 27. februar 2014 02.29.47 UTC+1 skrev snipped-for-privacy@attt.bizz:

yes when they run off the same clock, but if you use multiple ADC each with their own PLL to do sample rates that isn't a nice fraction of the xtal rate I'm not so sure

-Lasse

I understand

could you occasionally feed all the inputs from a calibration source? sorta like a clapper board

if you want to get fancy something like a prbs to get a nice correlation peak

must it be done real time?

-Lasse

Yes, that is possible but painful because it requires extra muxes.

Not really but close. If there is a delay of a few seconds that's ok but there will be signals coming in all the time, continuously.

What exactly are you trying to achieve ?

Create a 20 channel audio mixer ?

Sampling some 20 discrete channels ? At what bandwidth (sampling rate) and at what accuracy (number of bits or SNR) do you need ?

If you are doing something else than some multichannel audio application, AC97 might not be the first choice.

At least one or two decade ago the sound cards were really bad, with grossly incorrect sampling rate and bad temperature stability. Also samples were frequently lost.

I tried to make a Loran-C interference eliminator (noise blanker) for the 135 kHz band. To do so, I would have to generate clean synchronized GRIs for each Loran-C station, so I could mute the 135 kHz receiver, each time there was a Loran-C pulse from some stations.

In initial tests, since I live quite far from a Loran-C chain, I asked someone living closer to the Loran-C to send me some Loran-C recordings at 100 kHz. While a local oscillator or sound card sampling error would be quite easy to compensate, lost samples and sampling rate wobble made it quite impossible to generate a clean version of the GRI, thus making the GRI noise blanker quite useless.

Perhaps the sound cards are better now, but since the Loran-C is obsolete, this is a non-issue.

Can't reveal details but it's 20+ electrical signals that spectrally happen to be in the audio range. Several kHz. Sampling range needs to be

Afterwards the data gets heavily number-crunched in the PC.

I already did something similar with AC97 but that only needed two input channels and one output channel. Now I need more, a lot more.

I found sound card performance stellar so far, even for the ones on my older Samsung NC10 or the Durabook 14. Using the Daqarta tool kit I was able to reliably and most of all repeatably measure phase shifts of fractions of a degree.

I think if you want to do it for a reasonable price, it's going to go that way.

My point was that depending on the power level needed, separate DACs and ADCs can be easier than CODECs (keeping the ADCs away from power amps).